Self-supervised learning (SSL) techniques have achieved remarkable results in various speech processing tasks. Nonetheless, a significant challenge remains in reducing the reliance on vast amounts of speech data for pre-training. This paper proposes to address this challenge by leveraging synthetic speech to augment a low-resource pre-training corpus. We construct a high-quality text-to-speech (TTS) system with limited resources using SSL features and generate a large synthetic corpus for pre-training. Experimental results demonstrate that our proposed approach effectively reduces the demand for speech data by 90% with only slight performance degradation. To the best of our knowledge, this is the first work aiming to enhance low-resource self-supervised learning in speech processing.
@article{arxiv.2309.17020,
title = {Low-Resource Self-Supervised Learning with SSL-Enhanced TTS},
author = {Po-chun Hsu and Ali Elkahky and Wei-Ning Hsu and Yossi Adi and Tu Anh Nguyen and Jade Copet and Emmanuel Dupoux and Hung-yi Lee and Abdelrahman Mohamed},
journal= {arXiv preprint arXiv:2309.17020},
year = {2024}
}